Variability of Sediment Removal in a Semi-Arid Watershed William L

University of South Carolina Scholar Commons

Faculty Publications Geography, Department of

1983 Variability of Sediment Removal in a Semi-Arid Watershed William L. Graf University of South Carolina- Columbia, [email protected]

Follow this and additional works at: https://scholarcommons.sc.edu/geog_facpub Part of the Geography Commons

This Article is brought to you by the Geography, Department of at Scholar Commons. It has been accepted for inclusion in Faculty Publications by an authorized administrator of Scholar Commons. For more information, please contact [email protected]. WATER RESOURCES RESEARCH, VOL. 19, NO. 3, PAGES 643-652, JUNE 1983

Variability of SedimentRemoval in a SemiaridWatershed

WILLIAM L. GRAF

Departmentof Geographyand Centerfor SouthwestStudies, Arizona State University,Tempe, Arizona 85287

Field and documentarydata from Walnut Gulch Watershed,an instrumentedsemiarid drainage basin of approximately150 km 2 (57 mi2) in southeasternArizona, show that 83% of thealluvium removed from the basinduring a 15-yearerosion episode beginning about 1930 was excavatedfrom the highest-order stream.The amount of alluvium removedin the erosionepisode would have beenequal to a coveringof about4 cm (1.6 in) overthe entirebasin. The rate of sedimentremoval during the erosionepisode was 18 timesgreater than the rate of presentchannel sediment transport. Production of sedimentfrom slopesand channelthroughput at presentrates are approximatelyequal, and refillingwill not occurunder present conditions.The channelforms left by the massiveevacuation of sedimentimpose controls on the spatial distributionof tractiveforce and total streampower that make renewedstorage of sedimentlikely in only a few restricted locations. Modern instrumented records of a decade or more provide an inadequate perspectiveon long-termsediment movement.

INTRODUCTION omorphic province [Hunt, 1974]. Geologic materials include The geomorphologistand engineershare a common interest Precambrian volcanics and tertiary alluvium [Gilluly, 1956; in the transport and storage of sediment in semiarid water- Wilson, 1962]. Elevation rangesfrom about 1220 m (4000 ft) to sheds.Using theoreticaldeductions, the engineerhas frequently about 1890 m (6200 ft). The climate is semiarid with mean attempted to predict the behavior of sedimenttransport sys- annual precipitationof about 35.6 cm (14 in) and mean annual tems,while the geomorphologisthas frequentlyused field ob- temperature of about 17øC (63øF), with wide ranges for both servationsto inductively arrive at explanationsfor the process variables [Sellers and Hill, 1974]. Precipitation falls mostly as [Shen, 1979, p. 20/10]. The following paper representsan at- rain from winter stormsor suddensummer thunderstorms, so tempt to combine engineeringand geomorphologicalper- that the entrenchedchannels in the basinare usuallydry. The spectiveson the sedimenttransport processes in channelsof a vegetationin the basinis grass,shrubs, and brush. semiarid watershed, Walnut Gulch in southern Arizona. Sev- Human impacts in the basin include the townsite of Tomb- day pro- stone and associatedprecious metal mines begun in 1877, well doc n [Meyers,1956]. The mineshafts provided little surficialdistur- the watershed[e,g., Renardand Laursen,1975]. The emphasis bance,but the wastematerials from the shaftsand from milling inthe presentpaper is on basinwide considerationsof almosta operations,active until about 1930, impactedlimited areas of centuryof processoperation. the landscape.A railroad servingthe mining area and numer- There are three basic researchquestions regarding the various cross-basinroads may have affected channel processes, ation in sediment transportation in the Walnut Gulch Water- especiallythrough stabilizingthe channelgradient by harden- shed. First, how has sedimenttransport varied through time? ing road/channelintersections. In 1953 the U.S. Department of Short-termstudies have providedestimates of the magnitudeof Agriculturebegan an intensiveinstrumentation program in the presentsediment transport processes on slopesand in channels basin which included the installation of numerous concrete whichmay be placedin a contextwhen compared with century- flumes in the channel system [see Ferreira, 1979]. The flumes long record. Second,what has been the spatial distribution of resultedin further stabilization of establishedstream gradients. sedimentremoval during the last erosionepisode when arroyo Cattle grazing has affectedvegetation cover and related hydro- developmentoccurred? Significant channel erosion and con- 1ogic/geomorphicprocesses since Spanish incursions from comitant sedimentremoval occurred in the watershed during Mexico more than two centuriesago. the major erosionepisode that spannedmost of the southwest- Stream channels of Walnut Gulch watershed were mostly narrow and shallow in the late 1800's and meandered across the ern United States [Cooke and Reeves,1976]. Third, what has that erosion implied for continued sediment transport? The upper surfacesof alluvial fills. Plat maps drawn in 1881 by the channel morphology left by the catastrophicerosion episode General Land Office Survey show cienegas(wide grassymea -j representsa geometry that controls presently operating pro- dows) in a number of flow areas, especiallyat junctions of cesses[Cooke and Reeves,1976]. valleysin the center of the basin. In the lower reachesof the trunk stream, the channel was wide, shallow, and sandy. Be- STUDY AREA ginning in about 1930, residentsof the area report that the The Walnut Gulch Watershed is located in southern Arizona channelsof the entire drainage basin were entrenchedand the on pediment gravels,limestone, and igneousoutcrops near the cienegaswere destroyed. Massive amounts of sediment were town of Tombstone (Figure 1). The drainage area above the evacuatedfrom the basin, as occurredin many streamsof the lowest U.S. Department of Agriculture measurementsite is semiarid and arid southwestern United States [Cooke and about 150 km2 (57.7mi2). The rollingterrain of the basinhas Reeves,1976]. The entrenchedchannels remained in 1981, with sandy and gravelly soils typical of the Basin and Range ge- small amountsof refilling,as observedelsewhere in the Ameri- can southwest[Emmett, 1974; Leopold,1976].

Paper number 3W0283. Analysis of sediment removal during the erosion episode 0043-1397/83/003W-0283 $05.00 dependson knowledge of changing channel dimensions.The

643 644 GRAF'VARIABILITY OF SEDIMENT REMOVAL

USA ARIZONA

PhOenix

/nut Gulch Intensi•

Tombstone

I II Am 3

Fig. 1. Walnut Gulch Watershed, southernArizona.

dimensionsin turn provide the basisfor the analysisof sedi- throughoutthe basin.For an analysisof the channelsof the ment volumes removed during the erosion episode. Field entire basin,the data were organizedaccording to the Strahler checksshow that channelerosion during the past centuryis not stream order method [Doornkamp and King, 1971]. The likely to have incisedbedrock but rather has excavatedalluvial channelorders used were from a previousstudy by Murphy et deposits.Therefore, if the dimensionsof the small,pre-erosion al. [ 1977],permitting the calculationof the amountof sediment channelsand of the enlargedpost-erosion channels are known removedfrom an averagecross section for eachorder. The total at various cross sections, the differences between areas under amount of sedimenteroded from the system(which is seventh the two crosssections represent the amountsof materialeroded order)was represented by the function. (Figure 2). Data for channel cross sectionsfrom the post- erosionperiod were available from 43 surveyedcross-channel S,- Y',/_...q. (1) profilesmade in 1961 along a 13-km (8-mi) reach of'the main stem of Walnut Gulch. The profile data are stored at the where SouthwestRangeland Watershed Research Center in Tucson, S, total volume of sedimentremoved from the basinchannel Arizona. Data for the channel cross sections from the pre- system; erosion period were from the General Land Office Survey u stream order; recordsmade in 1881 and 1905, with the majority for the later n maximum order; date. Plat maps and surveyor'snotes are availablein the Phoe- L, total lengthof streamsof order u; nix, Arizona, office of the Bureau of Land Management for 55 S, mean crosssectional area of sedimentremoved by erosion crosssections scattered throughout the basin. Although more on streams of order u. data would be useful, the historical record is limited. The cross-sectionaldata for sediment removed by erosion Calculation of the total volume of sediment removed from were extended into the third dimension along channels the highestorder streamwas possiblein greaterdetail than an overallaverage because of numerouscross sections on the main stream. The total reach was divided into 43 subunits,with each subunit centeredon a measuredcross section (see Figure 1 for the location of the reach of intensive study). The amount of sediment removed from the total reach then was "

= Y, (2) $=1 where

S, total sedimentvolume eroded from the seventh(highest) 1961Surface • Bedrock'f½,'/½• Alluvium order channel; s number ofcross section' .... 1905Surface • AlluviumRemoved, 1905-1961 m maximum number ofcross section(43 in this case); Fig. 2. Schematicdiagram showinga typical crosssection on the S• crosssectional area of sedimentremoved from crosssec- main streamof Walnut Gulch and illustratingthe methodof determin- tion s' ing sedimentvolume eroded from the crosssection. Not to scale. L• lengthof the subunitcentered on crosssection s. GRAF: VARIABILITY OF SEDIMENT REMOVAL 645

Ds-1 Ds Ds+l where•:c is criticaltractive force in Newtonsper meterssquared and d is the medianparticle diameter in millimeters.The impor- Do I I I tance of critical tractive force to the present study is that S-1 S S+1 tractive force generatedby some flows may be insufficientto move commonly found particle sizesin some reachesof the I I channel. Da Db RESERVATIONS Ls The resultsof the investigationof spatialand temporalvari- Fig. 3. Schematic diagram defining factors used to calculate ation of sedimenttransport must be viewed with certain con- volume of sedimenttransported during the erosionepisode on Walnut Gulch. straintsin mind. The channeldimensions for the period before the erosionepisode are from the notesof surveyorswho were more concerned with township survey than with stream channels.General Land Office surveysthat can be checkedin other areas by independentevidence have proven reliable in Each of the surveyedcross sections had two typesof data' Ss, some casesbut not in others [Cooke and Reeves, 1976]. The cross-sectionalarea of lost sedimentand D s, downstreamdis- recordsin the Walnut Gulch area appear reasonablein light of tance of the cross section from an arbitrary starting point photographicevidence and recollectionsof long-time residents (Figure 3). The boundariesof eachsection were located halfway [Hastingsand Turner, 1965]. betweenthe measuredcross sections, so that the length of the Also, the original surveysdo not provide samplesof channel end subunitsrequired for the solutionof (2) was dimensionsas frequentlyalong the main stream a• do later D2--D • surveys.The lack of depth information at many crosssections L• = D• + (3) 2 in the older surveysrequired the use of depth dimensions calculated from allometric relationships established from Dm -- Dm- 1 L m= (4) known crosssections [Bull, 1975; Graf, 1979], further reducing 2 the precisionof the data. Channel dimensiondata from the old surveysrepresent a definiteimprovement over qualitative esti- Two measuressupply potential processinformation in the mates, however. following analyses.First is the conceptof 'tractive force' sug- The estimateof sedimentremoved during the erosionepisode gested,named, and definedby DuBoys in 1879 [Bogardi, 1978, involved comparisonof the pre-erosionchannels with those p. 80] where surveyed in 1961. Most of the erosion probably occurred % = 7RS (5) during eventsof intenseprecipitation in the early 1930'sand early 1940's(precipitation data publishedby Cookeand Reeves where [1976, pp. 65-79]). An implicit assumptionwas made that after z0 meantractive force, N m-2, the episodeof erosionwas completed(by 1945 at the latest),no 7 unitweight of water(9807 N m- 3); channelenlargement occurred. This assumptionwas probably R mean hydraulicradius,m; not strictly met, and some bank or bed scour probably oc- S energyslope (must be assumedto be equal to bed slopein curred after the erosionepisode and before 1961. The amount field applications). of material involved, however,is not likely to have altered the calculationsappreciably because there were few significantpre- It is perhapsmisleading to referto the DuBoys value as tractive cipitation episodesduring the period [see Cooke and Reeves, force sinceit is not a force in the physical-mathematicalsense, 1976,pp. 70-72]. but the term and equationare widely accepted[Bogardi, 1978, In the comparisonof rates of process,average yearly values p. 83]. The geomorphicsignificance of tractive forceis that it is were used for convenience,but actually the processesare dis- directly related to the competenceof stream flow over a wide continuous,and year to year variation is substantial.Along the range of particle sizes (recent reviews by Baker and Ritter channel,periods of scourand fill alternatewith eachother. The [1965] and Church[1978]). The units for tractive force(e.g., as average rates of sediment yield from slopesdo not take into shownby Church[1978, p. 756]) are the sameas the units for account climatic variability but submergesuch variability in shearstress (as given by Stelczer[1981, p. 18]). long-term averages.Sediment production rates are likely to The second measure of potential processis total stream have been significantly different in the period prior to the power, the amount of power expendedby flowing water per erosion episode when the vegetation cover was different in unit lengthof channel.As definedby Bagnold[1966, 1977] it is many parts of the basin.Variation in human-inducedchanges, f• = 7QS = %VW = 7RSVW (6) such as road building, mining, and grazing of cattle, also limit the extension of estimates of production of sediment from where f• is total stream power in N, V is mean velocity of flow slopes. in m s-1, W is widthof flow in meters,and othersymbols as before. The geomorphicsignificance of total stream power is SPATIAL VARIATION that it is directly related to total sedimentdischarge, that is, The total amount of sediment removed from the channels stream capacity assuminga constant supply of sediment(see was enoughmaterial to cover the basin to a depth of 4 cm (1.6 GRAF [1971] for a review). in). The valuesreported in Table 1 showthat the major portion Church [1978] found that the critical tractive force for very of sediment removed from the channels of Walnut Gulch loose sediment was Watershed was removed from the highest-orderchannel. Des- •:c= 1.78 d (7) pite the great numbersof lower-order channels,the relatively 646 GRAF' VARIABILITYOF SEDIMENTREMOVAL

TABLE 1. SedimentVolumes Removed During ErosionEpisode, Walnut Gulch Watershed,Southern Arizona

Mean Mean Mean Mean Stream Drainage Cross-Sectional Individual Total Volume, Stream Stream Length, Area, Area, Volume, Volume, % of Order Number m km2 m 2 m3 m 3 Total

1 6134 96 0.01 0.1 10 61,300 1 2 1373 190 0.03 0.2 38 52,200 1 3 257 508 0.13 0.8 406 104,000 2 4 58 1,306 0.50 2.4 3,134 182,000 3 5 15 2,430 1.95 7.3 17,739 266,000 4 6 3 5,712 7.62 22.2 126,806 380,000 6 7 1 13,051 146.52 * * 5,100,000 83 Total 6,150,000 100

Table values in the final column are rounded. *Mean valuesreplaced by detailed surveydata. small amount of material removed from each one results in low transport types and regions).In Walnut Gulch much of the amounts of sedimentbeing erodedfrom them in total. Whether sedimentis coarsecontinental alluvium carried by the streams or not theseresults are typical of semiarid watershedsremains as bedload.The particlesize d8,• for the alluvium is in the 4.0- to to be seen,because data are not availablefor comparison.In a 8.0-mm range [Osterkampet al., 1982]. Field investigationre- similar analysisof basinsin a humid region,J. C. Knox (Uni- vealedmany particles25 mm or larger in mediumdiameter. In versityof Wisconsin,Madison, personal communication) found lower ordersdepth of flow is insufficientto generatethe high that most of the sediment lost from channels in an erosion valuesof tractiveforce (or competence)required for transport. episodecame from the lower-orderstreams. The large amounts of sedimenteroded from the highest- The spatialvariation of erosionis directlyrelated in budget- order of the channelnetwork (as shownin Table 1 and Figure ary fashion to a comparisonbetween total streampower and 5) direct attention to a detailed analysisof the main streamof the amount of sedimentsupplied to the channelsystem. It is Walnut Gulch. A plot of crosssectional area of sedimentre- impossibleto construct a complete time seriesof sediment moved by erosion againstdownstream distance in the trunk supplyfrom surroundingslopes, but the channelsare excavated stream (Figure 6) showsthat the amount of material removed into•unconsolidated materials, either alluvium deposited by the first increasesand then decreasesdownstream. General expla- streamin its geologicallyrecent configuration or basinfill shed nation of the trends in Figure 6 dependson sedimentsupply, from Surroundingmountains. Therefore, throughout the period tractive force,and total streampower. of interestsediment supply has exceededtransport capacity, As statedpreviously, sediment supply has exceededtransport and spatialvariation in form is attributableto spatialvariation in transportcapacity or total streampower. 500.0 - The Walnut Gulch caseprobably differsfrom the southwestern Wisconsinexample because of the particlesizes involved. In Wisconsin,sediments are from fine-grainedloess soils (Figure 4) so that they are carried in suspensionand bedload is not a large proportion of the total load (Schumm[ 1977] reviewsriver 100.0

50.0 100-

ß-• 80 (o 10.0 -

• 5.0 ._o 60, Walnut Gulch ..... LoessWatershed, • "'o / i- 40- • 1.0 .g ßo- /'

0.5- 20-

0 0.01 0'.05 0:5 5'.0 0'.0 0.1, /i Grain Size, mm Order Fig. 4. Particle size analysesfor the channelsof the main trunk of Walnut Gulch and a typical loesswatershed in the midwesternUnited Fig. 5. Distribution by stream order of mean area of channelcross States. sectionexcavated by erosionin channelsof Walnut 6ulch Watershed. GRAF' VARIABILITY OF SEDIMENTREMOVAL 647

Distance Downstream, m

0 5,000 10,000 15,000 1200 I I .I

1000 -

-10,000 -1• o - (3,) E o (3,) 800 - E

(3,) - E .t 'O 600 - E

03 - -5,000 <• 400 < ._o o • 200 x i x

0 ' i I I I I 0 5,000 10,000 20,000 30,000 40,000 000

Distance Downstream, ft Fig. 6. Downstreamdistribution of area of channelcross section excavated by erosionon the main streamof Walnut Gulch. SeeFigure 1 for the locationof this intensivelystudied reach. capacity throughout the past century and is not a variable in removal include a peak of sediment removal near 4000 m this discussion.Generally, as higher orders are considered,the (about 14,000ft). Sedimentarymaterials in the reach are domi- 10-year dischargeat any one crosssection increases,channel nated by fine-grained, easily entrained materials, so that the slope exhibits relatively small declines, and width increases arroyo left after the erosionepisode was significantlywider and dramatically (Figure 7). The interaction of dischargeand width, deeperin this reach than in nearby reaches.The wide fluctu- however,causes a decreasein depth from sixth to seventhorder ations in amount of material removed from the lowermost (Figures 6 and 7). The amount of material excavatedfrom the reachesresult from the influencesof geologicstructure and the upper reachesof the network is relatively small where depth of nature of the sediment transport processitself. In some in- flow is insufficient to entrain the basin materials. In midbasin stances, areas where little erosion occurred had stored little areas,tractive force (competence)is sufficientto entrain materi- sedimentinitially becausethe channel was eroding through als and the channel is wide enough to permit high values of exposedresistant strata. In such places the channel was rela- total power (capacity),so large amounts of material were lost. tively narrow, deep, and had a steep gradient. Other fluctu- In the lowest reaches,the channelbecame so wide that depth of ations in the amount of sediment removed from the lowermost flow and tractive force declined and limited the amount of reachesreflect massesor pulsesof sedimentsmoving through material lost by erosion.Transmission losses, unaccounted for the system.Because the sedimentsare moved by discontinuous in this analysis,also depleted available dischargeand tractive events,these pulses are probably temporary features. force,as Wellas total streampower [Renard, 1970; Lane et al., 1980]. TEMPORAL VARIATION The final distribution of erosion is modified by a variety of The amount of material eroded from channel storageareas local influencesextending a few hundredsof meters along the during the erosionepisode can be placedin a temporal context channels in a variety of locations. These local phenomena by comparisonwith data collectedin other studiesof Walnut representvariability in bank resistanceso that the generaltrend Gulch. Comparisonsare possiblefor channel storage on a of changing amounts of channel erosion in the downstream several-thousand-yearscale and on a scaleof a few yearsduring direction, as shown in Figure 6, is imperfect.At downstream the post-erosionperiod for sedimentfrom slopesand channels. distance8230 m (27,000ft), a minimum of erosionindicates that Although large volumes of materials are involved in the relativelysmall amountsof material were removedin compari- recenterosion episode (about 6.2 x 106 m3), the volumeis sonto adjacentchannel areas. The area of the reducedsediment small in comparison to the amount of alluvium remaining loss is spatially coincidentwith surfaceoutcrops of Schieffelin beneath the channels. A well in the lower reach of the main Granodiorite (Tertiary) and Naco Limestone(Permian). These stream channel revealedthat alluvium extendedto a depth of highly resistantmaterials cause more narrow, deep, and steep 29 m (95 ft) below the surfaceof the modern channel [Renard, channelmorphology than the Tertiary-Quaternaryalluvial ma- 1977]. Therefore, based on survey data of channel crosssec- terials that dominate the other reachesof the main stream (for a tions at a similar site nearby and assumingcontinuity of sub- geologicmap of the basin, see U.S. Departmentof Agriculture surfaceconditions, the episodeof erosion removed only about [1970]). Channelsthrough the resistantsection contained less 15% of the alluvial materials stored in the lower reaches. Even alluvium for potential erosionthan other nearby reaches. allowing for substantialentrenchment of the channeland lesser Other specificdetails of the spatialdistribution of sediment depths of alluvium upstream,the amount of material removed 648 GRAF' VARIABILITY OF SEDIMENT REMOVAL

100.0-

50.0-

10.0- ./ 2.0-

5.0 1.0- E

0.5-

1.0

0.5 .•'

Order

0.1

Order

200.0

100.0

E 5o.o 0.020- (• ß

0.010- ._•

2 0.005- ø,,- 10.0 ./.

0.001

Order 1.0

Order

Fig. 7. The distribution by stream order of mean values of hydraulic parametersin channels of Walnut Gulch Watershed.A, width' B, depth'C, channelbed slope' D, magnitudeof calculated10-year peak discharge. in the erosionepisode accounted for only a minor portion of sedimentremoval during the erosionepisode therefore was 18 the total amount still stored in the basin. timesthe rate observedafter the episode. Local accounts indicate that most of the catastrophic Sedimentcontributions to the channelsfrom surrounding channelerosion occurred within a 15-yearperiod commencing slopesunder conditionsof the post-erosionperiod are about about 1930.The erosionprocess was inconsistent from one year equal to the amount of material transported through the to the next, but the overall averagerate of sedimenttransport channels. In recent work, Sireantonet al. [1977, 1980] and from channelstorage was about 434,000 m 3 per year.Studies Renard [1980] have calculatedthat 20,000 to 25,000 m3 of by Renard [1972] and Renard and Laursen [1975] show that material are erodedfrom the basin slopes.Given the accuracy during the post-erosionperiod the yearly averageof sediment of the estimatesand variablelengths of record,the annualslope output from the entirebasin was about 24,100m 3. The rate of erosion can be considered identical to the annual channel GRAF' VARIABILITYOF SEDIMENTREMOVAL 649

Distance Downstream, m 5,000 10,000 15,000 I I I. lOOO

900 -

800 -

700 200

.,-, 600

"'" 500 • 400 100 3oo

2oo

lOO

0 i i I i -0 0 10,•)00 20,000 30,•)00 ' 40,•)00 ' 50,000 Distance Downstream, ft

Downstream Distance, m 5,000 10,000 15,000 i I • -1.5

E b

o 101000 ' 20,•00 ' 30,•)00 ' 40,•)00 50,000 Downstream Distance, ft

Distance Downstream, m o 5,000 lO,OOO 15,000 I I I

150 E

lOO c•

I I I I I I i I o 0 10,000 20,000 30,000 40,000 50,000 Distance Downstream, ft Fig. 8. Downstreamdistribution of hydraulicvariables in the main trunk of Walnut GulchWatershed. A, width' B, depth;C, magnitudeof the 10-yearpeak discharge. Channel bed slope relatively invariate and not plotted. transportof 24,100m 3. If the presentrates of sedimentpro- arid streamsof Walnut Gulch are transportedas bedload and ductionfrom slopesand throughoutwere to be maintained,the becausein the main channellarge amountsof sedimentsremain material excavatedfrom the basin during the erosion episode available for entrainment,tractive force and total streampower would not be replaced. representreasonable indicators of the ability of the channelsto transport sediment [Graf, 1971]. When these measuresare IMPLICATIONS FOR PRESENT PROCESSES calculatedfor the dischargeof the 10-yearflood along Walnut The spatial distributionof sedimentremoval left behind Gulch (flood data from previouswork, publishedin part by channelforms that have importantimplications for continuing Kniselet al. [1979] and Reichand Renard[1981]), the resulting processes.Because the vast majority of sedimentsin the semi- distributionsshow that presentconditions are different from 650 GRAF: VARIABILITY OF SEDIMENT REMOVAL

20,000- indication that the coarse particles are not entrained in the lower reaches.The large particlesrequire almost 45 N m2 10,000- TotalStream Power/ .... Tractive Force tractive forcefor motion (from (7) and 25-mm particles),and as shown in. Figure 9, this critical level is not attained in some 5,000 reaches. In the main trunk of Walnut Gulch, downstream trends in z tractive force are much differentafter the erosionperiod than they were before (Figure 10). An erratic but generaldecline in tractive force in the downstreamdirection replaceda marked 1,000 downstreamincrease. It appearslikely that the largestparticles in the stream sedimentscan be carried in the upper reachesbut 500- 5OO not the lower reachesof the trunk stream,while the reversewas true before arroyo development. In the trunk, total streampower increasedin the downstream E direction more rapidly before the erosion episode than after z (Figure 11). The post-erosionsystem lacks the steep gradients 100- lOO• in total power found in the pre-erosion system, resulting in o more throughput of sedimentsin midbasinarea at present.In 50- 5o the more recent channel system,well-defined boxlike channels .> (which replaced wide shallow swales)produce higher levels of (,..3 total power in the upper reachesof the main stream because before arroyo developmentthe channelcould not contain all the water in the 10-yearflow. 10- 10 One of the most important implicationsof the spatial variation of sedimentremoval is that it imposesa particular spatial Order control on subsequentfluvial processes.The channelmorphol- Fig. 9. Distributions by stream order of mean values for tractive ogy left after the erosion episode dictates the likely loci of forceand total streampower for channelsof Walnut Gulch Watershed. erosion and deposition. Since sediment is readily available, those reacheswith sharply declining tractive force and total power are likely depositionsites, while those with sharp in- those prior to the erosion episode(see Graf [1982] for a de- creasesare likely erosion sites. The placement of one or a tailed analysis from a different area). Figures 7, 8, and 9 show limited number of recording instrumentsis therefore likely to the spatialdistribution of hydraulicmeasures. provide an inaccurateor misleadingrepresentation of the true Under conditions of the post-erosionperiod, total stream nature of processesalong the main stream.The temporal vari- power increasesconsistently with increasingorder (Figure 9). ation of sediment removal also makes limited recorded data Tractive force also increaseswith increasing order until the difficult to interpret, sincesuch data are usefulin a predictive highestor seventhorder is encountered(Figure 9). Dramatic senseonly until the next catastrophicadjustment. increasesin width and concomitant declines in depth serve to The calculationsreported here rely on flow generatedby the limit tractive force and competence(Figure 7). The largest 10-yearflood, which has a magnitudetoo small to substantially particlesin the channelmay thereforebe carried in midbasin affectthe channelmorphology of the network within the span areas, but in lower reachesthey may not be transported be- of a singleevent. The continuedmovement of materials from cause of lower levels of tractive force. Renard and Laursen low-order streamsand storage in high-order streamsmight in [1975] noted the fine characteristicsof sedimenttransported in time alter the spatial arrangementsof tractive force and total the lower reaches as opposed to upstream areas, a further stream power, but the role of high-magnitudeevents in such a

Downstream Distance, m 5,000 10,000 15 000 120 I I 1905 100 1961 8O

0 ' 10,•)00 ' 20,•)00 ' 30,(•00 ' 40,•)00 50,000 Downstream Distance, ft Fig. 10. Downstreamdistributions of tractiveforce for the main trunk of Walnut Gulch. SeeFigure 1 for locationof this intensivelystudied reach. For referencepurposes the largestcommon particles have critical tractive force levels of about 45 Nm-2 GRAF: VARIABILITY OF SEDIMENT REMOVAL 651

Distance Downstream, m o 5000 10,000 15,000 25,000 I I I

• Z 20,000

-• O•15,000 '• Q"10,000 '•E / x•/ .... 1905 / • 1961 ! 0- I I I I 0 10,000 20,000 30,000 40,000 50 000

Distance Downstream, ft Fig. 11. Downstreamdistributions of total streampower for the main trunk of Walnut Gulch. SeeFigure 1 for location of thisintensively studied reach.

schemeis unknown. The 100-year event, for example, might without the assistance of several individuals at the research center. Ken completelychange the presentsystem and establishnew spatial Renard, ResearchDirector, and Leonard Lane, hydrologist,cooper- ated during the project and offered helpful suggestionsin the prep- arrangementsover a shortperiod. aration of this report. Del Wallace, geologist,generously provided valuabledata for the study. CONCLUSIONS Tentative answers are possible for the basic research questions.First, rates of sedimenttransport during the erosion REFERENCES episode were about 18 times greater than channel sediment Bagnold, R. A., An approach to the sedimenttransport problem from transport rates or production of sedimentfrom slopesduring generalphysics, U.S. Geol.Surv. Prof. Pap. 422-J, 21 pp., 1966. the post-erosionperiod. The precipitation record [Cooke and Bagnold, R. A., Bed load transport by natural rivers, Water Resour. Reeves,1976, pp. 70-73] showsat leasttwo remarkableyears of Res., 13, 303-312, 1977. intense precipitation during the episode.Second, during the Baker, V. R., and D. F. Ritter, Competenceof riversto transportcoarse bedload material, Geol. Soc. Am. Bull., 86, 975-978, 1975. erosion episode beginning about 1930 the majority of sedi- Bull, W. B., Allometric changeof landforms,Geol. Soc. Am. Bull., 86, ments evacuated from the basin were eroded from the highest- 1489-1498, 1975. order stream.At crosssections along the length of the highest- Bogardi,J., SedimentTransport in Alluvial Streams,Akademiai Kiado, order segment,the amount of sedimentremoved increasedin Budapest,1978. Church, M., Paleohydrologicalreconstructions from a Holocenevalley the downstream direction to a maximum and then declined, a fill, in Fluvial Sedimentology,edited by A.D. Miall, pp. 743-772, distributionreflecting the conflictinginfluences of variation in Canadian Societyof PetroleumGeologists, Calgary, 1978. dischargeand channelwidth. Third, the arroyo forms left by Cooke, R. U., and R. W. Reeves,Arroyos and EnvironmentalChange in the erosion episodecontrol the distribution of tractive force the American South-West, Clarendon, Oxford, 1976. and total streampower so that renewedstorage of sedimentis Doornkamp, J. C., and C. A.M. King, Numerical Analysisin Ge- omorphology,St. Martin's Press,New York, 1971. likely in only a few limited areas. Emmett, W. W., Channel aggradation in western United States as A geomorphologicalperspective on the sedimenttranspor- indicated by observationat Vigil Network sites.Z. Geomorphol.21, tation processeson the Walnut Gulch watershedas outlined 52-62, 1974. above indicatesthat short-termanalyses of the processespro- Ferreira, V. A., Bibliography: SouthwestRangeland Watershed Re- search Center, report, 51 pp., U.S. Dep. of Agric., Tucson, Ariz., duceaccurate but impreciseviews on the processesthat are not 1979. comprehensive.Process rates obtained from instrumentedpor- Gilluly, J., General geology of central Cochise County, Arizona, U.S. tions of the basin during the past three decadescannot be Geol.Surv. Prof. Pap. 281, 169 pp., 1956. extended spatially or temporally without risk of substantial Graf, W. H., Hydraulics of SedimentTransport, McGraw-Hill, New York, 1971. error. Without the short-term analyses,the longer-term ge- Graf, W. L., Developmentof montane arroyos and gullies,Earth Surf omorphologic perspectiveprovides a highly generalizedview Proc., 4, 1-14, 1979. lacking in detail, especiallywith respectto presentlyoperating Graf, W. L., Downstreamchanges in streampower, Henry Mountains, processes.Evidence from the Walnut Gulch Watershed sug- Utah, Ann. Assoc.Am. Geographers,in press,1983. geststhat the prudent watershedanalyst is one who employsa Hastings,J. R., and R. M. Turner, The ChangingMile, University of Arizona Press,Tucson, Arizona, 1965. judicious mixture of engineeringand geomorphicapproaches Hunt, C. B., Natural Regionsof the United States and Canada, W. H. to basicand appliedresearch. Freeman, San Francisco, 1974. Knisel, W. G., K. G. Renard, and L. J. Lane, Hydrologic data col- Acknowledgments.I am indebted to Victor R. Baker (Department lection: How long is enough?,in Proceedingsfrom the Specialty of Geosciences,University of Arizona) for indispensibleassistance in Conferenceon Irrigation and Drainage in the Nineteen-eighties,pp. unravelingthe tangledliterature pertaining to streampower. Much of 238-254, American Societyof Civil Engineers,Albuquerque, New the background developmentfor the researchreported in this paper Mexico, 1979. was supported by grant EAR-7727698 from the National Science Lane, L. J., V. A. Ferreira, and E. D. Shirley, Estimating transmission Foundation,Earth SciencesDivision, Geology Program.The research loss in ephemeralstream channels,H ydrol. Water Resour. Ariz. on Walnut Gulch was financiallysupported by a grant from the U.S. Southwest,10, 193-202, 1980. Department of Agriculture for a cooperativeagreement between the Leopold, L. B., Reversalof erosion cycle and climatic change,Quat. SouthwesternRangeland Watershed ResearchCenter (Agricultural Res., 6, 557-562, 1976. ResearchService, Tucson) and the Department of Geography,Arizona Leopold, L. B., and W. W. Emmett, Bedloadmeasurements, East Fork State University. The research could not have been accomplished River, Wyoming, Proc. Nat. Acad. Sci., 73, 1000-1004, 1976. 652 GRAF: VARIABILITY OF SEDIMENT REMOVAL

Meyers, J., The Last Chance: Tombstone'sEarly Years, Dutton, New Sellers,W. D., and R. H. Hill, Arizona Climate, 1931-1972, University York, 1950. of Arizona Press,Tucson, 1974. Murphy, J. B., D. E. Wallace, and L. J. Lane, Geomorphicparameters Schumm, S. A., The Fluvial System,Wiley-Interscience, New York, predict hydrographcharacteristics in the Southwest,Water Resour. 1977. Bull., 13, 25-38, 1977. Shen,H. W., Modelingof Rivers,John Wiley, New York, 1979. Osterkamp, W. R., E. R. Hedman, and A. G. Wiseman, Geometry, Simanton, J. R., H. B. Osborn, and K. G. Renard, Effects of brush to basin characteristics,discharge, and particle-sizedata from gaged grassconversion on the hydrology and erosion of a semi-aridsouth- stream-channelsites, Western United States, U.S. Geol. Surv. Open western rangeland watershed,Hydrol. Water Resour.Ariz. South- File Rep. 82-93, 56 pp., 1982. west, 7, 249-256, 1977. Reich, B. M., and K. G. Renard, Application of advancesin flood Simanton, J. R., H. B. Osborn, and K. G. Renard, Application of the frequencyanalysis, Water Resour.Bull., 17, 67-74, 1981. USLE to southwesternrangelands, Hydrol. Water Resour. Ariz. Renard,K. G., The hydrologyof semiaridrangeland watersheds, A•lric. Southwest,10, 213-220, 1980. Res.Serv. Rep. ARS 41-162, 26 pp., 1970. Stelczer, K., Bed-Load Transport: Theory and Practice, Water Re- Renard, K. G., Dynamic structureof ephemeralstreams, Ph.D. disser- sourcesPublications, Littleton, Colorado, 1981. tation, Dep. of Civ. Eng. and Eng. Mech., Univ. of Ariz., Tucson, U.S. Department of Agriculture, Soil survey, Walnut Gulch experi- 1972. mental watershed,Arizona, spec.rep., 55 pp., U.S. Dep. of Agric., Renard, K. G., Past, present,and future water resourcesresearch in Tucson, Ariz., 1970. arid and semiarid areas of the southwesternUnited States, paper Wilson, E. D., A resumeof the geology of Arizona, Ariz. Bur. Mines presentedat the Hydrology Symposium,Inst. of Eng., Brisbane, Bull. 171, 140 pp., 1962. Australia, June 28-30, 1977. Renard, K. G., Estimating erosion and sedimentyield from rangeland, paper presentedat the Symposiumon WatershedManagement, Am. Soc.of Civ. Eng., Boise,Idaho, 1980. (ReceivedAugust 4, 1982; Renard, K. G., and E. M. Laursen, Dynamic behavior of ephemeral revisedFebruary 1, 1983; streams,,I. Hydraul. Div. Am. Soc.Civ. Eng., 101, 511-528, 1975. acceptedFebruary 16, 1983.)